Linux发展史

Published on 2016 - 07 - 19

Free-flowing UNIX culture at Bell Labs

From the very beginning, the UNIX operating system was created and nurtured in a communal environment. Its creation was not driven by market needs, but by a desire to overcome impediments to producing programs. AT&T, which owned the UNIX trademark originally, eventually made UNIX into a commercial product, but by that time, many of the concepts (and even much of the early code) that made UNIX special had fallen into the public domain.

If you are not old enough to remember when AT&T split up in 1984, you may not remember a time when AT&T was “the” phone company. Up until the early 1980s, AT&T didn't have to think much about competition because if you wanted a phone in the United States, you had to go to AT&T. It had the luxury of funding pure research projects. The mecca for such projects was the Bell Laboratories site in Murray Hill, New Jersey.

After a project called Multics failed in around 1969, Bell Labs employees Ken Thompson and Dennis Ritchie set off on their own to create an operating system that would offer an improved environment for developing software.

The simplicity and power of the UNIX design began breaking down barriers that, until this point, had impeded software developers. The foundation of UNIX was set with several key elements:

  • The UNIX filesystem—Because it included a structure that allowed levels of subdirectories (which, for today's desktop users, looks like folders inside folders), UNIX could be used to organize the files and directories in intuitive ways. Furthermore, complex methods of accessing disks, tapes, and other devices were greatly simplified by representing those devices as individual device files that you could also access as items in a directory.
  • Input/output redirection—Early UNIX systems also included input redirection and pipes. From a command line, UNIX users could direct the output of a command to a file using a right-arrow key (>). Later, the concept of pipes (|) was added where the output of one command could be directed to the input of another command. For example, the following command line concatenates (cat) file1 and file2, sorts (sort) the lines in those files alphabetically, paginates the sorted text for printing (pr), and directs the output to the computer's default printer (lpr):$ cat file1 file2 | sort | pr | lpr
    This method of directing input and output enabled developers to create their own specialized utilities that could be joined with existing utilities. This modularity made it possible for lots of code to be developed by lots of different people. A user could just put together the pieces as needed.

  • Portability—Simplifying the experience of using UNIX also led to it becoming extraordinarily portable to run on different computers. By having device drivers (represented by files in the filesystem tree), UNIX could present an interface to applications in such a way that the programs didn't have to know about the details of the underlying hardware. To later port UNIX to another system, developers had only to change the drivers. The application programs didn't have to change for different hardware!

To make portability a reality, however, a high-level programming language was needed to implement the software needed. To that end, Brian Kernighan and Dennis Ritchie created the C programming language. In 1973, UNIX was rewritten in C. Today, C is still the primary language used to create the UNIX (and Linux) operating system kernels.

Commercialized UNIX

Before the AT&T divestiture in 1984, when it was split up into AT&T and seven “Baby Bell” companies, AT&T was forbidden to sell computer systems. Companies that would later become Verizon, Qwest, and Alcatel-Lucent were all part of AT&T. As a result of AT&T's monopoly of the telephone system, the U.S. government was concerned that an unrestricted AT&T might dominate the fledgling computer industry.

Because AT&T was restricted from selling computers directly to customers before its divestiture, UNIX source code was licensed to universities for a nominal fee. There was no UNIX operating system for sale from AT&T that you didn't have to compile yourself.

Berkeley Software Distribution arrives

In 1975, UNIX V6 became the first version of UNIX available for widespread use outside Bell Laboratories. From this early UNIX source code, the first major variant of UNIX was created at University of California at Berkeley. It was named the Berkeley Software Distribution (BSD).

For most of the next decade, the BSD and Bell Labs versions of UNIX headed off in separate directions. BSD continued forward in the free-flowing, share-the-code manner that was the hallmark of the early Bell Labs UNIX, whereas AT&T started steering UNIX toward commercialization. With the formation of a separate UNIX Laboratory, which moved out of Murray Hill and down the road to Summit, New Jersey, AT&T began its attempts to commercialize UNIX. By 1984, divestiture was behind AT&T and it was ready to really start selling UNIX.

UNIX Laboratory and commercialization

The UNIX Laboratory was considered a jewel that couldn't quite find a home or a way to make a profit. As it moved between Bell Laboratories and other areas of AT&T, its name changed several times. It is probably best remembered by the name it had as it began its spin-off from AT&T: UNIX System Laboratories (USL).

The UNIX source code that came out of USL, the legacy of which was sold in part to Santa Cruz Operation (SCO), was used for a time as the basis for ever-dwindling lawsuits by SCO against major Linux vendors (such as IBM and Red Hat, Inc.). Because of that, I think the efforts from USL that have contributed to the success of Linux are lost on most people.

During the 1980s, of course, many computer companies were afraid that a newly divested AT&T would pose more of a threat to controlling the computer industry than would an upstart company in Redmond, Washington. To calm the fears of IBM, Intel, Digital Equipment Corporation, and other computer companies, the UNIX Lab made the following commitments to ensure a level playing field:

  • Source code only—Instead of producing its own boxed set of UNIX, AT&T continued to sell only source code and to make it available equally to all licensees. Each company would then port UNIX to its own equipment. It wasn't until about 1992, when the lab was spun off as a joint venture with Novell (called Univel), and then eventually sold to Novell, that a commercial boxed set of UNIX (called UnixWare) was produced directly from that source code.
  • Published interfaces—To create an environment of fairness and community to its OEMs (original equipment manufacturers), AT&T began standardizing what different ports of UNIX had to be able to do to still be called UNIX. To that end, Portable Operating System Interface (POSIX) standards and the AT&T UNIX System V Interface Definition (SVID) were specifications UNIX vendors could use to create compliant UNIX systems. Those same documents also served as road maps for the creation of Linux.
  • Technical approach—Again, until the very end of USL, most decisions on the direction of UNIX were made based on technical considerations. Management was promoted up through the technical ranks, and to my knowledge, there was never any talk of writing software to break other companies' software or otherwise restrict the success of USL's partners.

When USL eventually started taking on marketing experts and creating a desktop UNIX product for end users, Microsoft Windows already had a firm grasp on the desktop market. Also, because the direction of UNIX had always been toward source-code licensing destined for large computing systems, USL had pricing difficulties for its products. For example, on software that it was including with UNIX, USL found itself having to pay out per-computer licensing fees that were based on $100,000 mainframes instead of $2,000 PCs. Add to that the fact that no application programs were available with UnixWare, and you can see why the endeavor failed.

Successful marketing of UNIX systems at the time, however, was happening with other computer companies. SCO had found a niche market, primarily selling PC versions of UNIX running dumb terminals in small offices. Sun Microsystems was selling lots of UNIX workstations (originally based on BSD but merged with UNIX in SVR4) for programmers and high-end technology applications (such as stock trading).

Other commercial UNIX systems were also emerging by the 1980s as well. This new ownership assertion of UNIX was beginning to take its toll on the spirit of open contributions. Lawsuits were being initiated to protect UNIX source code and trademarks. In 1984, this new, restrictive UNIX gave rise to an organization that eventually led a path to Linux: the Free Software Foundation.

GNU transitions UNIX to freedom

In 1984, Richard M. Stallman started the GNU project (http://www.gnu.org), recursively named by the phrase GNU is Not UNIX. As a project of the Free Software Foundation (FSF), GNU was intended to become a recoding of the entire UNIX operating system that could be freely distributed.

Although rewriting millions of lines of code might seem daunting for one or two people, spreading the effort across dozens or even hundreds of programmers made the project possible. Remember that UNIX was designed to be built in separate pieces that could be piped together. Because they were reproducing commands and utilities with well-known, published interfaces, that effort could easily be split among many developers.

It turned out that not only could the same results be gained by all new code, but in some cases, that code was better than the original UNIX versions. Because everyone could see the code being produced for the project, poorly written code could be corrected quickly or replaced over time.

Over time, the term free software has been mostly replaced by the term open source software. The term “free software” is preferred by the Free Software Foundation, while open source software is promoted by the Open Source Initiative (http://www.opensource.org).

To accommodate both camps, some people use the term Free and Open Source Software (FOSS) instead. An underlying principle of FOSS, however, is that, although you are free to use the software as you like, you have some responsibility to make the improvements you make to the code available to others. In that way, everyone in the community can benefit from your work as you have benefited from the work of others.

To clearly define how open source software should be handled, the GNU software project created the GNU Public License, or GPL. Although many other software licenses cover slightly different approaches to protecting free software, the GPL is the most well known—and it's the one that covers the Linux kernel itself. Basic features of the GNU Public License include the following:

  • Author rights—The original author retains the rights to his or her software.
  • Free distribution—People can use the GNU software in their own software, changing and redistributing it as they please. They do, however, have to include the source code with their distribution (or make it easily available).
  • Copyright maintained—Even if you were to repackage and resell the software, the original GNU agreement must be maintained with the software, which means all future recipients of the software have the opportunity to change the source code, just as you did.

There is no warranty on GNU software. If something goes wrong, the original developer of the software has no obligation to fix the problem. However, many organizations, big and small, offer paid support (often in subscription form) for the software when it is included in their Linux or other open source software distribution.

BSD loses some steam

The one software project that had a chance of beating out Linux to be the premier open source kernel was the venerable BSD project. By the late 1980s, BSD developers at University of California (UC) Berkeley realized that they had already rewritten most of the UNIX source code they had received a decade earlier.

In 1989, UC Berkeley distributed its own UNIX-like code as Net/1 and later (in 1991) as Net/2. Just as UC Berkeley was preparing a complete, UNIX-like operating system that was free from all AT&T code, AT&T hit them with a lawsuit in 1992. The suit claimed that the software was written using trade secrets taken from AT&T's UNIX system.

It's important to note here that BSD developers had completely rewritten the copyright- protected code from AT&T. Copyright was the primary means AT&T used to protect its rights to the UNIX code. Some believe that if AT&T had patented the concepts covered in that code, there might not be a Linux (or any UNIX clone) operating system today.

The lawsuit was dropped when Novell bought UNIX System Laboratories from AT&T in 1994. But, during that critical period, there was enough fear and doubt about the legality of the BSD code that the momentum BSD had gained to that point in the fledgling open source community was lost. Many people started looking for another open source alternative. The time was ripe for a college student from Finland who was working on his own kernel.

Today, BSD versions are available from three major projects: FreeBSD, NetBSD, and OpenBSD. People generally characterize FreeBSD as the easiest to use, NetBSD as available on the most computer hardware platforms, and OpenBSD as fanatically secure. Many security-minded individuals still prefer BSD to Linux. Also, because of its licensing, BSD code can be used by proprietary software vendors, such as Microsoft and Apple, who don't want to share their operating system code with others. Mac OS X is built on a BSD derivative.

Linus builds the missing piece

Linus Torvalds started work on Linux in 1991, while he was a student at the University of Helsinki, Finland. He wanted to create a UNIX-like kernel so that he could use the same kind of operating system on his home PC that he used at school. At the time, Linus was using Minix, but he wanted to go beyond what the Minix standards permitted.

As noted earlier, Linus announced the first public version of the Linux kernel to the comp.os.minix newsgroup on August 25, 1991, although Torvalds guesses that the first version didn't actually come out until mid-September of that year.

Although Torvalds stated that Linux was written for the 386 processor and probably wasn't portable, others persisted in encouraging (and contributing to) a more portable approach in the early versions of Linux. By October 5, Linux 0.02 was released with much of the original assembly code rewritten in the C programming language, which made it possible to start porting it to other machines.

The Linux kernel was the last—and the most important—piece of code that was needed to complete a whole UNIX-like operating system under the GPL. So, when people started putting together distributions, the name Linux and not GNU is what stuck. Some distributions such as Debian, however, refer to themselves as GNU/Linux distributions. (Not including GNU in the title or subtitle of a Linux operating system is also a matter of much public grumbling by some members of the GNU project. See http://www.gnu.org.)

Today, Linux can be described as an open source UNIX-like operating system that reflects a combination of SVID, POSIX, and BSD compliance. Linux continues to aim toward compliance with POSIX as well as with standards set by the owner of the UNIX trademark, The Open Group (http://www.unix.org).

The non-profit Open Source Development Labs, renamed the Linux Foundation after merging with the Free Standards Group (http://www.linuxfoundation.org), which employs Linus Torvalds, manages the direction today of Linux development efforts. Its sponsors list is like a Who's Who of commercial Linux system and application vendors, including IBM, Red Hat, SUSE, Oracle, HP, Dell, Computer Associates, Intel, Cisco Systems, and others. The Linux Foundation's primary charter is to protect and accelerate the growth of Linux by providing legal protection and software development standards for Linux developers.

Although much of the thrust of corporate Linux efforts is on corporate, enterprise computing, huge improvements are continuing in the desktop arena as well. The KDE and GNOME desktop environments continuously improve the Linux experience for casual users. Newer lightweight desktop environments such as Xfce and LXDE now offer efficient alternatives that today bring Linux to thousands of netbook owners.

OSI open source definition

Linux provides a platform that lets software developers change the operating system as they like and get a wide range of help creating the applications they need. One of the watchdogs of the open source movement is the Open Source Initiative (OSI, http://www.opensource.org).

Although the primary goal of open source software is to make source code available, other goals of open source software are also defined by OSI in its open source definition. Most of the following rules for acceptable open source licenses serve to protect the freedom and integrity of the open source code:

  • Free distribution—An open source license can't require a fee from anyone who resells the software.
  • Source code—The source code must be included with the software and there can be no restrictions on redistribution.
  • Derived works—The license must allow modification and redistribution of the code under the same terms.
  • Integrity of the author's source code—The license may require that those who use the source code remove the original project's name or version if they change the source code.
  • No discrimination against persons or groups—The license must allow all people to be equally eligible to use the source code.
  • No discrimination against fields of endeavor—The license can't restrict a project from using the source code because it is commercial or because it is associated with a field of endeavor that the software provider doesn't like.
  • Distribution of license—No additional license should be needed to use and redistribute the software.
  • License must not be specific to a product—The license can't restrict the source code to a particular software distribution.
  • License must not restrict other software—The license can't prevent someone from including the open source software on the same medium as non–open source software.
  • License must be technology-neutral—The license can't restrict methods in which the source code can be redistributed.

Open source licenses used by software development projects must meet these criteria to be accepted as open source software by OSI. About 70 different licenses are accepted by OSI to be used to label software as “OSI Certified Open Source Software.” In addition to the GPL, other popular OSI-approved licenses include:

  • LGPL—The GNU Lesser General Public License (LGPL) is often used for distributing libraries that other application programs depend upon.
  • BSD—The Berkeley Software Distribution License allows redistribution of source code, with the requirement that the source code keep the BSD copyright notice and not use the names of contributors to endorse or promote derived software without written permission. A major difference from GPL, however, is that BSD does not require people modifying the code to pass those changes on to the community. As a result, proprietary software vendors such as Apple and Microsoft have used BSD code in their own operating systems.
  • MIT—The MIT license is like the BSD license, except that it doesn't include the endorsement and promotion requirement.
  • Mozilla—The Mozilla license covers the use and redistribution of source code associated with the Firefox web browser and other software related to the Mozilla project (http://www.mozilla.org). It is a much longer license than the others just mentioned because it contains more definitions of how contributors and those reusing the source code should behave. This includes submitting a file of changes when submitting modifications and that those making their own additions to the code for redistribution should be aware of patent issues or other restrictions associated with their code.

The end result of open source code is software that has more flexibility to grow and fewer boundaries in how it can be used. Many believe that the fact that numerous people look over the source code for a project results in higher-quality software for everyone. As open source advocate Eric S. Raymond says in an often-quoted line, “Given enough eyeballs, all bugs are shallow.”

Reference